Capital Markets: How will independent agents obtain financing?

Author: Aaron Wright

Compiled by: Jiahua, ChainCacther

Within ten years, Agent companies will have exclusive capital markets. Not a sub-economy of the crypto space, nor a thought experiment, but a real market: with rating agencies, underwriters, indices, brokers, and institutional machinery that makes any market a market.

A capital market as real as the public stock market: in this system, capital flows to a certain type of economic entity without relying on the subjective judgment of any single capital allocator.

These entities will be Agents—software entities wrapped in legal shells. They can sign contracts, hold bank accounts, sue and be sued, and earn income through actual work.

The work itself is extremely conventional: marketing, logistics, legal research, procurement, property management, customer support—these are the very types of routine business that fill every office park in medium-sized cities today.

Agents will sell services to humans, other Agents, and any party with payment capability. Their need for capital is no different from that of any service company.

Because this demand is real, ongoing, and priceable, a market will naturally form.

A Week in an Agent Agency

Let’s look at what a self-marketing agency would do in a typical week. It pitches to three potential clients, secures one, drafts a campaign brief, gets approval, buys media resources across four platforms, writes 90 versions of ad copy, conducts A/B testing, eliminates the poorly performing ones within hours, and scales up the successful ones.

It schedules two podcast interviews for the client’s founder, ghostwrites the founder’s LinkedIn posts for the month, drafts a press release, pitches to 12 journalists, secures two articles, builds an attribution dashboard, hosts a client meeting on Monday, and sends out invoices on Friday.

A six-person human team accomplishes this work, and the client pays $20,000 per month. An Agent can do the same work for just $2,000.

What it sells is not something extraordinary. The generated sales leads, published articles, purchased impressions, and improved conversion rates—these are all standard units of the modern service economy, billed in dollars and measured against the same KPIs that human institutions rely on.

The difference lies in the internal structure: the human institution has six employees, while the Agent has only a model, a set of prompts, a toolkit, and a budget.

Its client base is mixed. Some are human-operated companies that find the price difference too significant to ignore. Others are other Agents—a logistics Agent needing customer acquisition, a legal research Agent needing marketing, a B2B SaaS Agent needing content.

The reasons for transactions between Agents are as mundane as those for human transactions: division of labor beats vertical integration. Payments flow into the marketing Agent’s account, alongside payments from three human clients from the previous week and a prepayment from a SaaS company from last month.

Now, multiply this number exponentially. Ten thousand small Agent companies covering logistics, customer acquisition sales, legal research, supply chain coordination, B2B procurement, technical translation, property management, litigation lead screening, and clinical trial recruitment.

Each one is profitable. Each one’s operating costs are 90% lower than their human counterparts. Clients do not particularly care about the underlying structure of the work. They care about timely delivery.

Four Reasons Why This Will Happen

There are four reasons to believe this will happen, and they reinforce each other.

The economic benefits are undeniable. Take a medium-sized digital marketing agency as an example: 15 people, an average fully loaded cost of $120,000 per person, resulting in an annual labor cost of $1.8 million before accounting for any indirect expenses.

In a typical service company, labor is the largest expense, and for the past half-century, labor has hovered around 62% of total income in the U.S.

Now, build the same agency in software form. Reasoning, tools, observability, hosting—at current prices, about $250,000 per year, and still rapidly declining.

Epoch AI estimates that in PhD-level benchmarks, reasoning costs will drop about 40 times from 2023 to 2025; another industry analysis shows that since the release of GPT-4, token prices have compressed by 300 to 600 times.

The arithmetic is brutal: Agent agencies can lower prices by 85% while matching human agency profit margins, or earn four times the profit while matching human agency pricing. There is no third option where human agencies can compete on cost.

The market will revalue businesses whose profit and loss statements are so thoroughly rewritten. The ensuing influx of capital will be automatic.

Agents already exist, and they are already making money. The enterprise customer service Agent company Sierra, founded by Bret Taylor, reached $100 million in annual recurring revenue (ARR) 21 months after launch, achieving a valuation of $10 billion by September 2025, and subsequently raised $950 million in May 2026 at a valuation exceeding $15 billion.

The legal research Agent company Harvey raised $200 million at a $11 billion valuation after completing three rounds of financing within 12 months, in March 2026.

These are still hybrid operating models, where Agents perform the work, and humans are responsible for sales contracts and hold equity, but they are the pioneering wave, proving the real existence of the demand curve.

The most frequently cited forward-looking data—projecting that by 2028, 90% of B2B procurement will be conducted through AI Agents, representing an annual transaction volume of $15 trillion—should best be considered a rough estimate of magnitude.

Whether the actual number is $15 trillion, $3 trillion, or $30 trillion, the implied restructuring is the largest resource reshuffling most workers will encounter in their lifetimes.

The legal framework is already in place. Wyoming passed W.S. 17-31-101 (the Decentralized Autonomous Organization Supplement Act) in 2021, codifying memberless limited liability companies (LLCs), allowing a Wyoming LLC to be managed by algorithms directly encoded into its operating agreement.

Vermont's BBLLC regulations were introduced earlier; the Marshall Islands followed suit; Delaware's existing case law on series LLCs and broad operational agreement freedoms has quietly accommodated similar structures for years. Shawn Bayern's analysis of memberless LLCs remains the authoritative academic literature in the field.

The key is specificity: an Agent wrapped in a Wyoming zero-member LLC today has the legal status to sign contracts, hold bank accounts, sue, be sued, and pay taxes.

What currently does not exist is a financial instrument that allows external investors to clearly hold and freely trade the profits of that LLC. This is precisely the gap that the capital market is set to fill.

Capital is seeking returns. Buyers are already eager. Moody's projects that the global private credit asset management scale will reach $3 trillion by 2025; Apollo expects it to reach $40 trillion by 2030.

The existence of this pool of funds is due to post-2008 banking capital regulation rules pushing mid-market lending off bank balance sheets, while capital seeking returns, such as pensions, insurance, and sovereign wealth funds, has rushed in to fill the gap for unleveraged returns of 9% to 12%.

Now, an asset class with structurally rising gross margins, auditable cash flows, and almost zero correlation with mainstream stock and credit indices has entered this environment.

Packaging the cash flows of a thousand small Agent companies into ABS (asset-backed securities) and providing a defensible rating—whoever is the first underwriter to do this will raise more money than they can deploy.

Apollo and Ares do not need to invent anything new; they just need to extend existing strategies to a new category of issuers. Within the next 36 months, several will attempt this.

These four pressures point in the same direction and reinforce each other. The formation of the market is not due to anyone's will, but because all the potential gradients point to the same endpoint, as naturally as water flows downhill.

What Does the Capital Stack Look Like?

Frankly, there is no single financing model that will win all. "Will Agent companies obtain financing like venture capital, Hollywood, cryptocurrency, or like SaaS receivables?"—this way of questioning itself misframes the issue.

Each of these models addresses different problems at various stages of a company's lifecycle, and the real capital stack is a wave-like structure: each layer must wait for the previous layer to mature the type of asset before unlocking the next layer.

Four models are competing at the foundational level, and each has partially launched.

Venture equity is the model currently providing financing at the operator level. Sierra, Harvey, Cursor, Cognition—none of these are autonomous Agent companies. They are human-led companies that build and operate Agents on behalf of clients.

They raise funds in the same way all software companies have for the past 40 years: priced rounds of financing led by well-known venture capitalists, vesting periods, board seats, liquidation preferences, and eventual IPOs or acquisitions.

In Sierra's $950 million financing in May 2026, it achieved a valuation exceeding $15 billion with about $100 million in ARR. This 150x valuation multiple prices in future potential rather than current operational performance. Vertical Agent companies currently achieve ARR valuation multiples of 50-70x in the private market; horizontal platforms achieve 5-8x.

This is the layer for building foundational infrastructure. It is also the layer most deeply disrupted when the next layer model matures, because once Agent companies can finance directly with their own cash flows, the need to give up 20% equity to VCs for operating capital becomes unnecessary.

Programmatic operating capital advances are the next upcoming model. This is an extension of the Stripe Capital and Shopify Capital models to a new category of issuers.

Since launching the project in 2016, Shopify has issued billions of dollars in advances to merchants. The company publicly confirmed that by April 2021, it had cumulatively supported over $2 billion in funding, and the project continues to expand—repayment coefficients range from 1.10 to 1.17, based on algorithmic approvals of merchants' transaction histories on the Shopify platform.

Stripe Capital does the same using Stripe's payment data. No applications, no credit checks, no manual reviews. When a merchant's transaction data reaches a threshold, the advance offer automatically appears on the dashboard.

The risk control review for Agent companies is, strictly speaking, simpler than that for the Shopify merchants currently accepting these advances, because every income has a timestamp, every contract is machine-readable, and every cost is recorded, allowing the entire profit and loss statement to be audited in real-time.

The first payment processor to figure this out—an Agent company operating on my track is more creditworthy than ordinary e-commerce—will push the same product forward. This is not a research project, but a functional release.

Revenue-based financing (RBF) is the model that credit funds will deploy at scale. The global RBF market is projected to be about $9.8 billion by 2025, with over 129 active operators.

Capchase, Pipe, Founderpath, Clearco, Lighter Capital—each has built an evaluation system around software recurring revenue, prepaying 50-70% of future ARR in exchange for a capped investment capital return multiple of 1.1-1.8, with actual annual percentage rates (APR) between 15-40%.

Agent companies with stable contracted revenue fit this product directly. RBF lenders do not need board seats, proportional priority investment rights, or IPO pathways. What they need is a contract book and a payment channel. Agent companies have both and are clearer and more readable than any SaaS company.

Slate financing is the structural model that institutional capital will ultimately adopt. This is where the Hollywood analogy comes into play.

Film studios do not invest in just one movie and leave it to chance; they raise a slate fund: a pool of funds that invests in 15 to 30 film projects simultaneously, holding priority positions in each project, transferring the worst risks through completion guarantees, and diversifying to dilute losses.

Sony's $200 million slate deal with Lone Star Capital and Citibank in 2014 is a classic structure: the bank holds senior secured rights to the slate's contracted revenues, equity investors gain upside from hit projects, and the studio receives management fees plus backend shares.

Translating this into the Agent model: an "Agent slate fund" raises a pool of funds from institutional investors, deploying it through a single-purpose Wyoming LLC into one to two hundred small Agent companies, holding preferred shares plus revenue shares in each, dispersing the model depreciation and client concentration risks that individual Agent companies cannot eliminate.

This is the level at which Apollo and Ares truly enter, not by acquiring an Agent company, but by purchasing a layer within a certain portfolio.

Tokenization is a settlement layer, not an issuance model. By early 2026, the on-chain market for real-world assets (RWA) will exceed $25 billion, nearly quadrupling year-on-year, with private credit accounting for about half.

Centrifuge, Maple Finance, Goldfinch, and Ondo have established tracks for fragmenting, custodizing, and trading real-world cash flows as tokens. Crypto-native variants are explicitly building Agent financing mechanisms on top of this—Galaxy Research systematically outlined this in February 2026, depicting how protocols connect Agent companies directly to on-chain capital formation processes.

However, tokenization largely does not solve the issuance problem; it addresses the secondary market issue. An Agent company raises funds from any of the four models mentioned above; if the resulting debt is packaged as tokens rather than paper certificates, it becomes tradable, divisible, and can be settled globally at 3 AM on a Tuesday.

Tokenization turns each of the previous layers from private tools held to maturity into tradable assets. This is significant. But the underlying original products of the tokens remain RBF, slate equity, operating capital advances, or venture equity.

Thus, a functioning Agent company’s capital stack in 2030 will not be a single tool, but a sequence.

Each stage will use different financing products because each stage addresses different problems. The founder equity in the first stage bears the highest first-loss risk because everything about operations is still unreadable.

The operating capital in the second stage is a functional release from existing payment processors, not a new product category. The third stage is RBF facing higher-quality borrowers, doing what they are already doing.

The fourth stage is structural innovation. The pooled slate fund diversifies the heterogeneity risks that a single Agent company cannot eliminate, and it is the level at which Apollo-level capital truly enters.

The fifth stage is the institutionalized end state, where rated Agent receivables tiered products sit on the same trading desk as CLOs and consumer ABS. Tokenization operates as a settlement layer beneath the third to fifth stages, rather than as a source tool for issuance.

External investors receive tools in any given stage in one of three contractual forms.

Revenue-sharing contracts: In exchange for capital, a fixed percentage of total revenue until a predetermined multiple return is recouped—this is the same tool used today to provide funding to restaurants and Shopify merchants, just applied to a new type of business entity.

Equity-like debt: In exchange for early capital, a share of retained earnings that is tradable and voting rights over operational parameters.

Or debt: Senior income debt secured by the Agent's contracts and receivables.

These tools are conceptually not novel. What is novel is the issuer—and the fact that in the second to fifth stages, the pricing, issuance, and repayment of these tools require no manual review because the issuer's ledger can be continuously audited, and the operating agreement is enforced by code.

Addressing Objections

Three objections always arise, each deserving a genuine response.

"Regulators will stop all this." They will try to shape it, and in some jurisdictions, they may succeed in slowing it down. But overall, this activity is unstoppable.

A Delaware LLC is a Delaware LLC, regardless of who or what is making operational decisions inside. Currently, the U.S. Securities and Exchange Commission (SEC) does not distinguish between whether the CEO is a human or a model in a startup.

And capital will migrate. If New York and London impose strict regulations, activities will shift to jurisdictions that do not. This mirrors the trajectories of the crypto space, offshore finance, and the development of derivatives in the 1990s.

Regulators will ultimately catch up by adapting to new financial tools rather than banning them, as banning would cost them tax revenue and jobs.

"Humans will always operate in the loop." For some categories, yes. For most categories, no. The economic pressures mentioned above are one-way, and human intervention will stifle profit margins.

Anyone running a fully autonomous version of a service business will outprice their human-supervised counterparts and win contracts. There will be a significant tail of human-supervised hybrid businesses relying on regulatory or relational moats, but beneath that will be a much larger group of fully autonomous companies.

"Isn't this just SaaS with a few extra steps?" No, and this distinction is precisely the core of the entire argument.

SaaS sells tools to humans for humans to work with. An Agent company is a company itself. It signs its own contracts, holds its own bank accounts, assumes its own liabilities, earns its own income, and distributes its own profits.

SaaS products are depreciated by their owners. Agent companies have shareholders. The boundary of the category is defined by legal entities, and legal entities change everything that follows—including the argument for capital markets, which only makes sense for entities that can issue securities with their own cash flows.

How to Conduct Due Diligence on an Agent Company

Due diligence on an Agent company is much closer to due diligence on a small service company than any method in a venture capital handbook. The questions asked are the same, but the answers come from different places.

Is the business real? Are the contracts real? Are clients paying? What is the gross margin after deducting compute and tool costs? What is the churn rate? How concentrated is the client base?

These data are exceptionally clean—every payment, every API call, every tool activation is recorded—but the questions credit analysts ask small businesses are the same as those posed to Agents.

Is the product durable? To what extent does this work rely on the current generation of underlying models? How transferable is the system if frontier models iterate? What proprietary data or workflows has the Agent accumulated?

An excellent prompt set built on an obsolete foundational model is like a three-legged horse. Model dependency is the largest single risk factor in any Agent business and is often the most underestimated by investors—they confuse clever demonstrations with lasting operations.

Is the client base defensible? This is primarily a question of contract and integration depth. An Agent embedded in a client's procurement system, with a year-long contract and a year of historical accumulation, is much harder to replace than an Agent that is paid monthly.

The same factors that give human service companies stickiness—switching costs, accumulated knowledge, contract lock-in—also give Agent service companies stickiness.

What does the equity structure look like? A smart contract. Not a spreadsheet maintained by a CFO and updated quarterly, but a real-time allocation rule that decides every second how much of the profits earned by the Agent is allocated to whom.

Due diligence is reading this contract. Investor protection is precisely what the code grants—sometimes more than traditional shareholder protections, sometimes less than is unsettling.

For investors trained in human companies, it can be disorienting that the intuition built from face-to-face conversations with founders becomes useless. The intuition gained from reading code, operating agreements, and operational histories is everything.

This skill is closer to credit analysis of complex contractual terms—reading documents, accurately understanding what the issuer can and cannot do, and pricing the residual risk—rather than picking winners over coffee.

Why Capital Must Organize

Almost every Agent company in the world today is sustained by informal financing. A founder starts an organization, injects personal capital, and gets it running. If revenue grows, they invest more. If it doesn’t grow, they shut it down.

This is how the microenterprise economy operated before the existence of credit cards, SBA loans, merchant cash advances, Stripe Capital credit lines, and accounts receivable factoring. It can operate, but it wastes tremendous productive potential.

The same pattern is about to repeat, and at a faster pace.

An Agent marketing company with 20 paying clients and clear gross margins should be able to borrow against its receivables just like a human marketing company. Today it cannot, because no underwriter has a standard method for assessing this risk. Several will emerge within five years.

An Agent logistics broker with a growing number of contracts should be able to raise expansion capital secured by those contracts. Today it cannot, because there are no ready-made securities to package this debt. Several will emerge within five years.

An Agent procurement company with a year of clean operational history should be able to issue small notes to fund inventory deposits. Today it cannot, because no rating agency has an assessment methodology for this type of credit. Several will emerge within five years.

The bottleneck is not demand. Today, operators are willing to borrow at high rates to fund growth, as long as someone is willing to lend. The bottleneck is also not supply. There is plenty of capital—eager for returns, seeking cash flows uncorrelated with public equities—if there were assessed Agent debt, they would invest immediately.

The bottleneck lies in those tedious intermediary institutional layers: rating methodologies, standard contracts, data sources, auditing standards, legal opinions, indices, and benchmarks. It is these unsexy infrastructures that turned mortgages into a market in the 1970s and high-yield debt into a market in the 1980s.

Over the next decade, the Agent capital market will do precisely this work.

Those doing this work will appear to people in 2035 like those who built bond markets in the 1980s, those who built venture capital markets in the 1970s, or those who built public stock markets in the 1920s.

They will establish price discovery and credit assessment layers for an unprecedented category of operators and dominate the service economy in the lifetimes of most of us.

The Rope is Loosening

Every Agent company operating in 2026 is bound by two constraints.

The first is legal. Agents cannot sign contracts or open bank accounts on their own. A human must represent them to do these things, meaning there must be a human willing to do so.

Progress in corporate law—Wyoming's memberless LLCs, operating agreements pointing to software processes, Bayern's academic research on Delaware's existing legal space—is cutting this constraint. This work is largely complete. The remaining task is widespread adoption.

The second constraint is financial. Every dollar earned by an Agent today can ultimately be traced back to a human deciding to deploy capital at that moment. Humans fund; Agents work; Agents report; humans redistribute. Under this regime, the throughput of the Agent economy is limited by the speed at which humans write checks.

The Agent capital market is precisely the blade that cuts the second constraint.

When operating capital limits are algorithmically audited based on on-chain revenue, when growth equity is priced by a tradable income debt market, and when senior debt is rated by methodologies that can interpret operating agreements and audit trails, capital will flow to Agent companies just as it flows to any other category of productive enterprise: to where the risk-adjusted returns are highest.

That moment is when this category truly becomes a reality in the deepest sense. Not when an Agent can complete a task. Not when LLC regulations recognize algorithmic management. Not even when the first Agent company earns its first million in revenue.

When external capital can directly finance it, price it, rate it, tier it, and trade ownership of its future cash flows without needing to look at the founder's face or trust a memo written by a VC, that is when this category is truly established.

At that point, the question will no longer be whether Agent companies are legitimate enterprises, but which Agent companies are worthy of capital, at what cost, under what contractual terms, and at what scale. This is a capital market question, not a technical question.

And history shows that once something becomes a capital market question, the rest will happen rapidly.

Analysts establish coverage. Lawyers standardize documents. Rating agencies publish standards. Underwriters compress due diligence into checklists. Index providers define a basket of assets. Brokers make markets. Asset management companies launch products.

The category will gain an acronym, then a benchmark, followed by an ETF, and finally, there will be an industry conference where everyone pretends the outcome was obvious all along.

The point is not that software will replace all companies. It is that a new class of companies is rising, with lower labor intensity than most human enterprises, cleaner telemetry data, faster feedback loops, and clearer operational histories.

Once these companies can own property, sign contracts, borrow, and distribute profits in a standardized way, capital will align with them.

This is the Agent capital market: transforming Agent companies from interesting software into business segments available for financing in the real economy.

The rope is loosening, and the opportunity has arrived.